CN106408520A - High-color fidelity image defogging method - Google Patents

Abstract

The invention discloses a high-color fidelity image defogging method. The method includes obtaining a group of new HSI imaging models in the frog weather by converting the image from an RGB space to an HSI space; performing defogging based on image enhancement on the image brightness component I in the HSI space and obtaining a new brightness component IJ through the combination with the HSI imaging models; performing modification by means of the new brightness component IJ and obtaining the saturation component SJ of the image through the combination with the HSI imaging models; and maintaining the tone component unchanged and converting the defogging result from an HSI image to an RGB image. Therefore, the defogged image has high color fidelity. The image is processed by the HSI space, so that the image color can be better maintained.

Description

A kind of image defogging method of high color fidelity

Technical field

The present invention relates to the mist elimination technical field based on image enhaucament is and in particular to a kind of image of high color fidelity goes Mist method.

Background technology

In the case of the greasy weather, imaging device is subject to scattering and the absorption of air suspended particles, and atmosphere light participates in imaging Impact, become picture contrast declines, and visibility reduces, and details is smudgy, and picture quality significantly declines.It is basic Imaging model is as follows:

I (x)=J (x) t (x)+A (1-t (x)) (20)

Wherein J (x) is script fog free images, and I (x) is Misty Image, and t (x) is the absorbance of scene, and A is atmospheric environment Light.Scene absorbance t (x) reflects the depth information of Misty Image, and t (x) is less to represent that the fog of impact image is thicker, is subject to Impact bigger.

The general mist elimination algorithm based on image enhaucament, is typically carried out in RGB color, by by R, G, B tri- Channel separation, and process separately as gray level image.But due in processing procedure, have ignored R, tri- passages of G, B completely Between contact it is easy to lead to the distortion of color.

Content of the invention

It is an object of the invention to overcoming deficiencies of the prior art, provide a kind of image of high color fidelity Defogging method.

The purpose of the present invention is achieved through the following technical solutions.

A kind of image defogging method of high color fidelity, comprises the steps：

(1) image is transformed into HSI space by rgb space, obtains the HSI imaging model under one group of new greasy weather；

(2) the luminance component I of image is carried out processing and combine HSI based on the mist elimination of image enhaucament in HSI space As model obtains new luminance component I_J；

(4) by new luminance component I_JIt is modified and obtains with reference to HSI imaging model the saturation component of image S_J；

(4) keep chrominance component constant, mist elimination result is converted to RGB image from HSI image.

Further, image by the change type in rgb space to HSI space is：

For the chrominance component H in HSI component, because fog can't impact to the tone of image script, So having：

H_I=H_J(24)

H here_IRepresent haze image tone, H_JRepresent fog free images tone.

But for component S saturation and I brightness, due to the impact of fog, saturation S significantly declines, and bright Degree I then rises overally, and contrast reduces, and details reduces.Next according to atmospherical scattering model come to mist figure and artwork Saturation, luminance component relation are derived.

First, the atmospherical scattering model of (1) formula is separated to R, tri- passages of G, B then have:

I_R(x)=R (x) t (x)+A (1-t (x)) (25)

I_G(x)=G (x) t (x)+A (1-t (x)) (26)

I_B(x)=B (x) t (x)+A (1-t (x)) (27)

It is assumed here that absorbance t (x) and atmospheric environment light A are to R, the impact of tri- passages of G, B is all identical.

By obtaining to (6), (7), (8) summation：

I_R+I_G+I_B=(R+G+B) t+A (1-t) (28)

In conjunction with (4) formula, this example can obtain：

I_I=I_Jt+A(1-t) (29)

Wherein I_IRepresent the luminance component of mist figure, I_JRepresent the luminance component of fogless figure.

Minimum Value Operations are carried out in R, G, B triple channel to (1) formula.

I_min(R,G,B)=J_min(R,G,B)t+A_min(R,G,B)(1-t) (30)

In conjunction with (1) formula and (11) formula, this example can obtain:

In fact, contrast (3), formula (12) left side is actual to be equal to saturation S_I, and to molecule denominator on the right of formula (12) Remove A simultaneously, just can obtain, and actually in the case of the greasy weather, this can be considered 0, therefore omit.Then formula (12) abbreviation is：

Formula (13) both sides are simultaneously except the further abbreviation of J is:

So just obtain the imaging model in HSI space in the case of the greasy weather:

In the case of formula (15) illustrates the greasy weather well, the imaging law of the image under HSI color space pattern.We can To process to (15) further, to obtain the form more suitable for image enhaucament.We are permissible for convolution (14) and formula (10) Eliminate absorbance t, obtain:

Wherein

Convolution (3) and formula (16) we can obtain:

Then we have just obtained the HSI imaging model under one group of new greasy weather:

So, we only need to the luminance graph of Misty Image is strengthened so that result just may be used close to min (R, G, B) To obtain by (16), (17) try to achieve I_JComponent and S_JComponent.

Further, by formula (18), the mist elimination algorithm of the image enhaucament in rgb space can be transformed into by we HSI space is processed.

We obtain the gray level image of the mist elimination of luminance component I by algorithm for image enhancement first, and using result as min (R, G, B) brings (17) formula into, and the result tried to achieve is as new luminance component I_J.Luminance component I due to such estimation_JIt is not The image of actual scene, is brought directly to the error that (16) formula is likely to result in required saturation, therefore bright to estimation here Degree component I_JIt is modified, introduce correction factor m and go correction formula (16) so as to the luminance component I that brings into_JDistribution value is more Close to actual distribution, so just can substantially estimate saturation S_J, and due to saturation S_JTo slight change and details letter Breath is simultaneously insensitive, and therefore after such process, we often can obtain the preferable result of color fidelity:

Due to the brightness I via image enhaucament_JOften bright than the luminance component under practical situation, therefore m generally takes Positive number, simultaneously in order to ensure that luminance component after treatment is not in the situation that data is overflowed, by its least commitment to 0.03.

Further, in process of the test, often assume that atmosphere light A perseverance is 1, repaiied also dependent on practical situation here Just, but in order to simplify algorithm, the present invention directly takes 1.

Compared with prior art, the invention has the advantages that and technique effect：

The present invention provides a kind of image defogging method of high color fidelity, is mainly based upon atmospherical scattering model to HSI Color space model is optimized, so that mist elimination image has higher color fidelity, the present invention adopts HSI space pair Image is processed, and can be very good to keep the color of image.The present invention effectively improves the fidelity of image, can obtain preferably Visual effect.

Brief description

The original haze image of Fig. 1 embodiment one.

Fig. 2 in RGB color model using partial histogram equalization obtained by mist elimination result figure.

The luminance component I that Fig. 3 is tried to achieve_J.

The saturation component S that Fig. 4 is tried to achieve_J.

The final process result of Fig. 5 embodiment one.

The original haze image of Fig. 6 embodiment two.

Fig. 7 in RGB color model using MSR obtained by mist elimination result.

The luminance component I that Fig. 8 is tried to achieve_J.

The saturation component S that Fig. 9 is tried to achieve_J.

The final process result of Figure 10 embodiment two.

Figure 11 is the schematic flow sheet of the image defogging method of high color fidelity.

Specific embodiment

To describe embodiments of the present invention below with reference to drawings and Examples in detail, thus elaborating on how to apply The HSI color space model of the present invention goes to optimize all kinds of image enhaucament mist elimination algorithms, and by using this HSI color space model The mist elimination algorithm of image enhaucament and the image enhaucament mist elimination algorithm in rgb space contrast, thus highlighting this color space mould The superiority of type.

Embodiment one

Example 1 adopts the mist elimination algorithm based on partial histogram equalization algorithm.Fig. 1 illustrates this example will be carried out The haze picture of mist elimination.This example first directly carries out mist elimination process to image in RGB color, and result is as shown in Fig. 2 permissible See that serious cross-color integrally in image.Next processed using the HSI color space model of this example, specifically Step is as follows：

1) Fig. 1 is converted into HSI figure；

2) to luminance component I_ICarry out partial histogram equalization process, and result is brought into (17) as min (R, G, B) Formula, the result tried to achieve is as new luminance component I_J, result is as shown in Figure 3；

3) by the I obtaining_J, bring formula (19) into thus trying to achieve saturation component S_J, result is as shown in Figure 4；

4) keep chrominance component constant, and result is converted back RGB image, result is as shown in Figure 5.

This example can be seen by being transformed into local histogram equalization algorithm in the HSI color space model of this example Go to process, can effectively improve the fidelity of image, obtain preferable visual effect.And for from run time, In HSI color space model, process time only has 51.5% (matlab R2015b, CPU i5-2410m@of RGB color 2.30GHz) it is seen that this algorithm not only has superior color fidelity, and can also effectively improve operational efficiency.

Embodiment two

Example 2 adopts the mist elimination algorithm based on MSR algorithm.Fig. 6 illustrates the haze figure of this example mist elimination to be carried out Piece.This example first directly carries out mist elimination process to image in RGB color, and result is as shown in Figure 7 it can be seen that image is overall Serious cross-color occurs.Next processed using the HSI color space model of this example, comprised the following steps that：

1) Fig. 6 is converted into HSI figure；

2) to luminance component I_ICarry out partial histogram equalization process, and result is brought into (17) as min (R, G, B) Formula, the result tried to achieve is as new luminance component I_J, result is as shown in Figure 8；

3) by the I obtaining_J, bring formula (19) into thus trying to achieve saturation component S_J, result is as shown in Figure 9；

4) keep chrominance component constant, and result is converted back RGB image, result is as shown in Figure 10.

Can see by MSR algorithm is transformed in the HSI color space model of this example go process, can be effective Improve the fidelity of image, obtain preferable visual effect.And for from run time, in HSI color space model Process time only has 49.8% (matlab R2015b, CPU i5-2410m@2.30GHz) of RGB color, more can see Obtain superiority and the versatility of this algorithm.Most of image enhaucament class mist elimination algorithm processing in rgb space can pass through This HSI color space model is processed to optimize original algorithm.

Although disclosed herein embodiment as above, described content is only to facilitate understanding the present invention and adopting Embodiment, is not limited to the present invention.Technical staff in any the technical field of the invention, without departing from this On the premise of the disclosed spirit and scope of invention, any modification and change can be made in the formal and details implemented, But the scope of patent protection of the present invention, still must be defined by the scope of which is defined in the appended claims.

Claims (5)

1. a kind of image defogging method of high color fidelity is it is characterised in that comprise the steps：

(1) image is transformed into HSI space by rgb space, obtains the HSI imaging model under one group of new greasy weather；

(2) the luminance component I of image is carried out processing and combine HSI imaging mould based on the mist elimination of image enhaucament in HSI space Type obtains new luminance component I_J；

(3) by new luminance component I_JIt is modified and obtains with reference to HSI imaging model the saturation component S of image_J；

(4) keep chrominance component constant, mist elimination result is converted to RGB image from HSI image.

2. a kind of image defogging method of high color fidelity according to claim 1 is it is characterised in that step (1) is concrete Including：

The basic imaging model of Misty Image is as follows:

I (x)=J (x) t (x)+A (1-t (x)) (1)

Wherein J (x) is script fog free images, and I (x) is Misty Image, and t (x) is the absorbance of scene, and A is atmospheric environment light；

Image by the change type in rgb space to HSI space is：

$H={\cos }^{-1}\left\{\frac{\frac{1}{2}\[(R-G)+(R-B)\]}{\sqrt{{(R-G)}^2+\left(R-B\right)\left(G-B\right)}}\right\}---\left(2\right)$

$S=1-\frac{3}{R+G+B}\min \left(R+G+B\right)---\left(3\right)$

$I=\frac{1}{3}\left(R+G+B\right)---\left(4\right)$

For the chrominance component H in HSI component, because fog can't impact to the tone of image script, have：

H_I=H_J(5)

H here_IRepresent haze image tone, H_JRepresent fog free images tone；

The atmospherical scattering model of (1) formula is separated to R, tri- passages of G, B then have:

I_R(x)=R (x) t (x)+A (1-t (x)) (6)

I_G(x)=G (x) t (x)+A (1-t (x)) (7)

I_B(x)=B (x) t (x)+A (1-t (x)) (8)

Assume absorbance t (x) and atmospheric environment light A to R, the impact of tri- passages of G, B is all identical；

By obtaining to (6), (7), (8) summation：

I_R+I_G+I_B=(RG+B) t+A (1-t) (9)

In conjunction with (4) formula, can obtain：

I_I=I_Jt+A(1-t) (10)

Wherein I_IRepresent the luminance component of mist figure, I_JRepresent the luminance component of fogless figure；

In R, G, B triple channel, minimum Value Operations are carried out to (1) formula,

I_min(R,G,B)=J_min(R,G,B)t+A_min(R,G,B)(1-t) (11)

In conjunction with (1) formula and (11) formula, obtain:

$\frac{I-3I_{\min \left(R,G,B\right)}}{I}=\frac{\left(J-3J_{\min \left(R,G,B\right)}\right)t+\left(A-3A_{\min \left(R,G,B\right)}\right)\left(1-t\right)}{Jt+A\left(1-t\right)}---\left(12\right)$

Formula (12) abbreviation is：

$S_I=\frac{(J-3J_{\min (R,G,B)})t}{I}---\left(13\right)$

Formula (13) both sides are simultaneously except the further abbreviation of J is:

$S_I=\frac{S_{J}Jt}{I}---\left(14\right)$

Obtain the imaging model in HSI space in the case of the greasy weather:

$\left\{\begin{array}{c}{H}_I=H_J\\ S_I=\frac{S_{J}Jt}{1}\\ I_I=I_{J}t+A(1-t)\end{array}\right.---\left(15\right)$

Convolution (14) and formula (10) eliminate absorbance t, obtain:

$S_J=k\left(\frac{A-J}{J}\right)---\left(16\right)$

Wherein

Convolution (3) and formula (16) can obtain:

$I_J=\frac{kA+min(R,G,B)}{1+k}---\left(17\right)$

Then obtained the HSI imaging model under one group of new greasy weather:

$\left\{\begin{array}{c}{H}_J=H_I\\ S_J=k\left(\frac{A-I_J}{I_J}\right)\\ I_J=\frac{kA+\min (R,G,B)}{1+k}\end{array}\right.---\left(18\right).$

3. the image defogging method of described a kind of high color fidelity according to claim 2 is it is characterised in that step (2) specifically include：

The mist elimination algorithm of the image enhaucament in rgb space can be transformed into HSI space and processed by formula (18)；Logical first Cross the gray level image that algorithm for image enhancement obtains the mist elimination of luminance component I, and result brought into (17) formula as min (R, G, B), The result tried to achieve is as new luminance component I_J.

4. the image defogging method of described a kind of high color fidelity according to claim 3 is it is characterised in that step (3) specifically include：

Due to new luminance component I_JIt is not the image of actual scene, (16) formula that is brought directly to is likely to result in required saturation Error, therefore here to estimation new luminance component I_JIt is modified, introduce a correction factor m and go correction formula (16) so as to the luminance component I bringing into_JDistribution value, closer to actual distribution, so just estimates saturation S_J, and due to Saturation S_JTo slight change and detailed information and insensitive, therefore can obtain the preferable result of color fidelity:

$S_J=k\left(\frac{A-max(0.03,(I_J-m\left)\right)}{\max (0.03,(I_J-m\left)\right)}\right)---\left(19\right)$

Due to the new luminance component I via image enhaucament_JOften bright than the luminance component under practical situation, therefore m is usual Take positive number.

5. the image defogging method of described a kind of high color fidelity according to claim 3 is it is characterised in that step (3) by m least commitment to 0.03 in.